Inequality grew when the economy shifted into hyperinflation in the mid- ... countries that have blatant economic disparities, such as Brazil and Guatemala.
The Rise of Economic Inequality Momi Dahan* * This work is composed of two parts. The first presents a current survey of changes in inequality in Israel, focusing on the factors that determine inequality in income distribution. Much research work has been invested in exploring the effect of government intervention on inequality. The effect or transfer payments and direct taxes has come under especially powerful scrutiny. We also discuss macroeconomic factors that affect inequality. Inequality in Israel (before government involvement is taken into account) is considerable and has been rising for almost three decades. The main cause of this economic inequality is inequality in the labor market. The second part of this study probes the trend in labor-market inequality. This important topic has not been adequately researched for the purpose of understanding the trend in inequality in Israel. This part of the chapter traces the origins of the upturn in labor-market inequality over the past two decades. It focuses on questions
*
I thank Leah Achdut, Avi Ben-Bassat, Elhanan Ben-Porat, Shlomo Yitzhaki,
Yoram Meshar, and Danny Tsiddon for their useful remarks. I also thank the participants in the Bank of Israel seminar for their useful remarks. Special gratitude is reserved for Gili Fernandez for his important contribution in preparing this study.
2
such as: Is the upturn in inequality attributable to a widening of schooling disparities? Surprisingly, the effect of the Israeli education system on economic inequality has not been researched. Has the return on schooling changed significantly? How has immigrant absorption affected the structure of return on human capital?
Part A: General Survey
1.
The Rising Tide of Economic Inequality
A glance at Figure 15.1, which covers the years 1965–1997, reveals a clear trend of widening economic disparities in Israeli society since the mid-1970s, as measured in terms of inequality in gross income among households headed by employees.1 There is a perceptible overlap between macroeconomic events and the trend in inequality. Inequality increased as the economy tumbled into a deep slump in 1967 and contracted immediately after the lethargy ended. Inequality grew when the economy shifted into hyperinflation in the mid-
1
The Gini index for gross-income inequality is an updated version of data in
Dahan (1995). Dahan (1995) presents a detailed survey of changes in definitions in the income surveys and their implications for the quality of comparisons over the years. One can probably use these data with all due caution. Unfortunately, there is no sufficiently lengthy series of inequality data for the other definitions of income.
3
1980s and contracted (although not fully) immediately after inflation was checked. Inequality ascended again (this time to record levels) with the advent of mass integration in the 1990s. Insert Figure 15.1 Table 15.1 presents inequality data for a shorter period, from 1979 to 1997, during which definitions were more uniform. The table shows that the economic inequality created by market forces, measured by a Gini index for inequality in economic income, has attained a troubling level and has been climbing steadily at the since 1980.2 Israel’s level of economic-income inequality, the effect of transfer payments and direct taxes is mechanically factored in, surpasses that of the United States (Table 15.6) and brings to mind countries that have blatant economic disparities, such as Brazil and Guatemala.
Table 15.1. Gini Index for Inequality among Households, Population at Large, 1979-1997
2
This remark does not refer to the perfect action of market forces. Economic-
income inequality is affected by the use of transfer payments and direct taxes (see discussion below). Moreover, the government provides public education that affects labor income.
4
Labor income
Economic income
Gross income
Net income
0.560
0.504
0.419
0.356
Private consumption N/A No survey No survey No survey No survey No survey No survey
1979 1980 1981
N/A 0.544 0.538
N/A 0.411 0.410
N/A 0.361 0.354
1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997
0.559
N/A N/A N/A N/A N/A N/A
0.413
0.352
N/A N/A
N/A N/A
No survey
No survey
No survey
No survey
0.328
0.559 0.555 0.568 0.576 0.586 0.585 0.579 0.588 0.579 0.586 0.585
0.498 0.496 0.513 0.516 0.524 0.529 0.526 0.535 0.527 0.532 0.533
0.414 0.408 0.421 0.417 0.423 0.432 0.422 0.440 0.432 0.432 0.432
0.361 0.358 0.367 0.366 0.373 0.378 0.368 0.385 0.372 0.375 0.372
No survey No survey No survey No survey No survey
N/A N/A
0.321 No survey No survey No survey No survey N/A
Despite the differences in data-gathering methods, demographic structure, and other attributes, an international comparison may be instructive. In the past year, the Latin American Development Bank has released figures on inequality in South American countries. They portray South America as the continent with the highest inequality indices in the world (IADB, 1999). According to this publication, the Gini index for per-capita gross-income inequality in South America ranged from 0.43 in Uruguay to 0.59 in Brazil. Since the rate of transfer payments in South American countries is negligible, these indices give an approximation of inequality in economic income. In comparison, in 1997
5
the Gini index for economic income inequality in Israel was 0.533 while inequality in labor income in Israel was even greater at 0.585 (Table 15.1). Clearly, the labor market “produces” considerable inequality, which is expressed in inequality in economic income. (The second part of this study is devoted entirely to the origins of the labor-market inequality.) Government involvement by means of transfer payments and taxes reduces economic disparities in Israeli society to the common European level. (For an expanded discussion, see the “International Comparison” section.)3 The Gini index for net-income inequality was 0.372 in 1997 and 0.333 after standard-person adjustment. (Notably, however, the inequality index for net income is flawed, as we show in the next section.) Inequality in standard of living, measured on the basis of the Gini index for private consumption, is even smaller than inequality in net income. Many tend to think that this measure provides the best summarizing indicator of standard-of-living inequality. Additionally, this index does not have the flaws that impair the net-income inequality index it. Inequality in well-being is even smaller after the calculus takes account of additional factors that affect
3
Thus far, I have not obtained data on economic-income inequality in any
European country. A comparison with a European country is warranted due to the similarity in extent of government involvement.
6
household well-being, such as number of children, amount of leisure time, and consumption of public services (Table 15.2).
Table 15.2. Distribution of Income and Standard of Living, Population at Large, 1992/1993 (percent) Decile Lowest 0.4
2
3
4
5
6
7
8
9
1.6 3.2 5.2 6.9 9.2 11.3 13.7 17.9 Econ. Income 2.1 3.7 5.0 5.8 6.9 9.1 10.7 12.9 15.7 Gross income 2.6 4.5 5.9 6.8 7.8 9.6 11.0 12.9 15.1 Net income 3.4 5.1 5.9 7.3 8.6 9.5 10.7 12.8 15.4 Priv. Consump tion 13.5 13.3 12.2 11.0 10.0 9.6 8.9 8.7 8.1 Educ. & healtha 1.3 1.3 2.0 1.9 1.9 2.2 2.1 2.1 2.0 No. of childrenb Leisure 96 87.4 84.5 80.3 79.5 76.1 76.0 73.5 71.4 hrs/weekb .9 a. The share of each decile in education and health services was computed on the basis of net
Upper most 30.6 28.1 23.8 21.2
4.7 2.2 70.5
income deciles per standard adult. b.
Number of children and hours of leisure time were computed among households in which the head of households was aged 30-44, in accordance with net income deciles of the population at large in the Income Survey (without adjustment for and standard adult).
Sources: income and private consumption—Central Bureau of Statistics, Household Expenditure Survey 1992/93. No adjustment for standard adult is made. Education and Health—Central Bureau of Statistics, Publication 1049 and Household Expenditure Survey 1992/1993.
Another difference stands out in Table 15.1: the long-term trend in economic inequality relative to net-income inequality. While economic-
7
income inequality has worsened considerably, net-income inequality has increased slightly only. The next section shows that the degree of transfer-payment intervention needed to offset the increase in economic inequality is rising. To reduce the inequality brought about by forces in the market, especially the labor market, high tax rates are needed to finance transfer payments for disadvantaged population groups. As we see below, such involvement may also be one of the factors in Israel’s severe economic inequality. 2.
Effect of Government Policy Measures on Inequality
Government intervention affects inequality in various ways that may be sorted into several categories: (a) the effect of the distribution of ownership of tangible assets, i.e., physical capital, for example, the sale of public housing at subsidized prices; (b) the effect of the distribution of intangible assets, i.e., human capital; for example, in the provision free public education; (c) the effect on the structure of return on tangible and intangible capital; for example, progressive income tax; (d) current public support, for example, old-age pensions; and (e) regulations and laws that do not have budget implications, for example, government regulations to protect monopolies. To the best of my knowledge, there is no information on the overall effect of government on the distribution of ownership of tangible assets. Even reasonable information about the distribution of capital in Israel is lacking. Unfortunately, this important topic is still shrouded in darkness. According to
8
conventional wisdom based on findings in other countries, capital is more unequally distributed than labor income. Due to the lack of the data on the distribution of capital, one cannot estimate, for example, the direct effect of equity grants on the distribution of capital in Israel. This is not the only effect; capital grants also affect employment and wages. The manner becomes even more complex when one bears in mind that foreign investors also benefit. Similarly, we cannot estimate the effect of government on net return on equity and the effect of monetary policy on inequality in society. The Household Expenditure Surveys of the Central Bureau of Statistics (CBS) include information on the distribution of ownership of housing, durable goods, and private savings.4 The Income Surveys also provide data on capital income.5 To the best of my knowledge, no study that utilizes these data to create an estimate of capital in Israel has been performed.
4
We should note that the private saving rate, according to the National
Accounts data, was 20.4 percent (on average for 1992 and 1993) while the saving rate implied by the Family Expenditure Survey for 1992/93 was 4.6 percent (Lavie and Spiwak, 1996). Although the private saving indicator shown in the National Accounts includes business saving, this does not seem sufficient to explain the entire discrepancy between the two sources of data. 5
When capital income in the Income Survey is compared with the share of
capital income in the National Accounts, large discrepancies appear.
9
Government action may affect inequality even if it has no budget implications. Government regulations that give a private entity a monopoly, e.g., in imports of electrical appliances, do not entail budget expenditure but benefit the monopoly holders, who receive a windfall by gouging consumers. In Israel, no comprehensive study has been performed that could estimate the impact of monopoly- and cartel-friendly regulations on inequality. a.
The Effect of Direct Taxes and Transfer Payments
However, the CBS surveys do provide reliable information on the distribution of ownership of human capital. Sufficiently reliable data also exist on income from labor, income from public support, and direct tax remittances. The current data in the Income Survey pertain only to households headed by
According to the Income Survey for 1995, the share of property income was 1.5 percent (of total labor, property, and pension income) and the share of pension income was 7.3 percent. In contrast, the rate of return on equity was 14.1 of in national income in 1995 (as opposed to a 22 percent rate of return on equity in business-sector product). Additionally, much of the government’s domestic debt is held by pension institutions. The share in GDP of government interest payments to the domestic public was 5.3 percent in 1995. There is no direct correspondence between property income to return on equity, but the immense discrepancy between these two sources points to a possible skewing in capital income as derived from the income surveys.
10
employees; income of the self-employed is not sufficiently reliable and is not published. The National Insurance Institute releases an annual survey on the effect of transfer payments and direct taxes on inequality among households headed by employees (and the unemployed), on the basis of data from the Household Income Survey. (The survey reports extensively on poverty and the effect of the policy measures on its incidence and intensity.) To estimate the effect of the policy measures, the authors of the survey calculate a Gini index for inequality in economic income (income from labor and capital), gross income (economic income plus transfer payments), and net income (gross income less direct taxes). As Table 15.3 shows, inequality is smaller in gross income per standard person than in economic income.6 In 1997, economic inequality among the population at large, expressed in Gini index terms, was 0.505 per standard person in respect to economic income. The inclusion of transfer payments in income lowered the inequality to 0.395, a 21.8 percent decrease. Inclusion of the effect of direct taxes lowered the index of net-income inequality by another 12.3 percent. These calculations show that transfer payments are much more effective than direct taxes in reducing inequality. Furthermore, the rate of contraction of inequality by transfer payments has been increasing over time.
6
Indices of inequality among households elicit similar results (Table 15.1).
11
This does not prove that the transfer payments are becoming more efficient, since the rate of transfer payments has also been rising commensurably. In direct taxes, too, one cannot point to such a clear trend, but the effect was greater in 1997 (the end of the period investigated) than in the early 1980s. In regard to taxes, too, these results imply nothing about efficacy. Practically speaking, this method does not provide information on the effectiveness of the policy measures.
Table 15.3. Effect of Transfer Payments and Direct Taxes on Inequality, Population at Large, 1979–1997
1979
0.432
0.366
0.318
Percent of decrease resulting from transfer payments 15.2
1980
0.434
0.369
0.324
14.9
10.4
1981
0.439
0.372
0.319
15.4
12.1
1982
0.444
0.367
0.312
17.3
12.4
1983
0.439
0.360
0.301
17.9
13.6
1984
0.472
0.398
0.327
15.8
15.1
1985
0.468
0.372
0.312
20.2
13.1
1986
No survey
No survey
No survey
No survey
No survey
1987
N/A
N/A
N/A
N/A
N/A
1988
0.457
0.370
0.322
19.1
10.5
1989
0.474
0.378
0.325
20.3
11.1
1990
0.480
0.376
0.326
21.7
10.3
1991
0.490
0.377
0.327
23.1
10.2
For economic incomea
Gini index For gross incomea
For net incomea
Percent of decrease resulting from direct taxes 11.1
12
1992
0.498
0.393
0.339
21.1
10.7
1993
0.494
0.383
0.329
22.5
10.9
1994
0.502
0.399
0.344
20.4
11.0
1995
0.497
0.397
0.337
20.2
12.1
1996
0.496
0.387
0.329
22.0
11.8
1997
0.505
0.395
0.333
21.8
12.3
a.
The Gini index for net income is adjusted to standard persons.
Source: National Insurance Institute, Annual Survey, 1997/98
I believe that the estimations of the effect of these policy measures on inequality, following the approach described above, are defective in several ways. Therefore, severe caution is warranted in making them available to decision-makers for use even as initial approximations. First, the data on transfer payments and direct taxes (unlike those on labor and capital income) are not obtained from interviews with respondents; instead, they are computed by officials at the Central Bureau of Statistics.7 Below is a quotation from a CBS publication that explains how to the data in the Household Income Survey are gathered and processed: . . . The material at the individual level is subjected to various computations and corrections, such as . , . computation of child allowances at the National Insurance rates; imputation of other
7
Other Western countries also behave similarly (Gottschalk and Smeeding,
1997).
13
benefits wherever reporting is incomplete; . . . computation of gross income for cases in which net income or a sum paid is reported; [and] computation of direct taxes on the basis of the income tax and National Insurance tables (CBS, Household Income Survey, 1992, 1993, Special Publication 99 3). Thus, the evaluation of the effect of policy measures on inequality assumes that direct taxes are collected “by the book” (and that households’ takeup of benefits is full) and are not based on income tax and National Insurance contributions actually paid and benefits actually received. The foregoing estimates of the effect of transfer payments and direct taxes are skewed insofar as tax evasion and incomplete takeup of entitlements are not factored into income. In the absence of information on the actual figures, the extent of the skew cannot be estimated. Second, the computation may be skewed even if all individuals pay taxes by the book (and if the CBS takes account all of the many provisions that affect the correct computation of taxes) and invoke their full eligibility for benefits. This is because this approach assumes, groundlessly, that individuals cannot adjust their labor in response to changes in tax rates and generosity of benefits. Finally, a budget action usually has implications on both the expenditure and the revenue sides of the budget. When we measure the effect of transfer payments on inequality using the method described above, we answer the
14
following question: how do transfer payments affect inequality if we disregard the way the budget sources are raised? This information is indeed important in assessing the relative efficiency of individual policy measure, but this method of measurement should not be used to assess the totality of a budget action (a “budget package”). Raising the rate of Value Added Tax would almost certainly be regressive if we ignore the uses to which the added revenues are put, but we might reach but the opposite conclusion if the added tax revenues were used to increase progressively distributed transfer payments. b.
The Effect of Direct Taxes on Inequality
Gabai (1996), investigating the effect of indirect taxes on income-distribution inequality, found that they have an upward influence on income inequality. However, the assessment of the impact of indirect taxes failed to take account of individuals’ responses to the imposition of taxes (how consumption patterns would change if the tax were to be rescinded). Additionally, since indirect taxes apply to products and not to individuals, and it is difficult to identify the bearer of the burden. The indirect-tax burden is apportioned between consumers and producers on the basis of the ratio of supply and demand elasticities for the product at issue. However, producers pay production factors that may be the very same consumers or the consumers of other products. This makes it even more complex to gauge the impact of indirect taxes on incomedistribution inequality.
15
In contrast, the effect of indirect taxes on the distribution of private consumption is very easy to estimate if the taxes are applied to all products at the same rate, as occurs in the case of Value Added Tax. One may state without thoroughgoing research that VAT is neutral in its effect on consumption inequality. Therefore, Israel’s changeover in the 1970s from multiple indirect-tax rates to a uniform rate of Value Added Tax, makes research on the effect of indirect taxes on inequality somewhat less necessary. After all, many regard private-consumption distribution as a better indicator than income distribution of inequality in standard of living. c.
Distribution of Public Services by Income Levels
Individuals also benefit from products in which government is involved in delivery, such as education, health, and social services. Therefore, in measuring the effect of government funding on financial-income inequality one must observe measures additional to those of financial nature, such as transfer payments and taxes. In a Special Publication of the Central Bureau of Statistics (1997) that did much to enrich the supply of information on the effect of government on inequality, the incremental in-kind income flowing from education and health services was estimated on the basis of income deciles.8
8
“Influence of In-Kind Transfers on Income Distribution,” Publication 1049.
16
The estimate of the financial value of education and health services included all services delivered at no charge or at reduced charge by central government, municipal government, National Institutions, and nonprofit organizations. The method of computation made use of the demographic makeup of households, per-pupil expenditure itemized by levels of education, and approximate health expenditure by age. In 1992, general government provided education and health services at a financial cost of NIS 14.6 billion (in 1992 prices), 17.5 percent of disposable household income. In comparison, NIS 15.2 billion was spent on transfer-payments that year. According to the estimate in Table 15.4, central government provides disadvantaged population groups with more education and health services, in terms of financial value, than it does to high income deciles. (The same is obviously true in terms of share of income.) The financial value of education services is 47 percent of net (mean) household income in the lowest decile as against only 9 percent of net household income in the uppermost decile. The government action is progressive in transfer payments, direct taxes, and— according to the foregoing estimate—the main social services (education and health). Economic income is twelve times higher in the uppermost decile than in the lowest decile. The ratio falls to approximately 5.3 when transfer payments and direct taxes income are factored in and to 3.7 when the financial value of education and health services is included. In fact, social services are distributed more equally than transfer payments are.
17
Table 15.4. Education and Health Services, by Net Income Decilesa (per standard adult), 1992/1993 Decile
Econ. income Transfers Direct taxes Net income Educ. and health a.
Lowest 1,020
2
3
4
5
6
7
8
9
Highest 12,541
2,051
2,703
3,514
3,996
5,097
6,186
7,654
9,628
919
913
829
748
678
610
523
489
516
625
58
178
270
464
531
766
1,077
1,514
2,243
3,223
1,881
2,786
3,262
3,768
4,143
4,941
5,632
6,629
7,901
9,943
880
868
798
715
653
626
578
569
525
306
Monthly household income (NIS)
Source: Central Bureau of Statistics (1997).
Although the breakdown of public-services distribution by income levels enhances our understanding of what the government does, the limitations of this yardstick should be kept in mind. First, the reckoning of the distribution of public expenditure for social services across income deciles assumed a 100 percent takeup rate. However, since the extent of takeup by the disadvantaged may not resemble that of the non-disadvantaged, the contraction of inequality shown in the computation may be overstated. Second, the act of adding financial income and the financial value of education and health systems presumes that this conjunction reflects the subjective value of these services to their recipients. However, the computation disregards the inherent utility of personal free choice in using one’s financial income, a factor that does not
18
exist when social services, although expressed in financial terms, are actually received in kind. Third, the financial value of education and health services was estimated on the basis of cost of delivery; if the services are delivered inefficiently, they may actually be worth less than the sum quoted.9 Finally, some public goods such as education, which are partly of investment nature, may eventually serve as a means of generating financial income. Differences in the quality of education services that different individuals receive may determine the financial return on their schooling, and this figure may not be equal to the cost of the services. In the second part of this chapter, we examine the contribution of education to financial income and the relationship between the distribution of schooling among the population and labor-wage inequality. d
Minimum Wage and Inequality
Another policy measure designed to reduce inequality among workers is the minimum wage. A collective agreement concluded between the Coordinating Bureau of Economic Organizations (“organized management”) and the Histadrut (General Federation of Labor) in 1972 assured every worker in Israel a minimum income pegged at 44 percent of the national average wage per employee post. Over the years, the minimum wage relative to the national
9
This problem recurs in measuring national product, which, of course, is the
sum of private consumption, investment, exports, and imports (negative sign), based on market prices, and public consumption on the basis of cost prices,
19
average wage eroded because it was adjusted mainly in tandem with Cost-ofLiving-Allowance agreements. In April 1987, this arrangement was solemnized in law and the minimum wage was set at 45 percent of the national average wage. The effect of the Minimum Wage Law on inequality, assuming full compliance with the law, is reflected in the bent trajectory of the left-hand tail of the distribution curve and in a change in the wage scale of persons who previously received a near-minimum wage. Dahan (1995) found no support (in the inequality time series) for the theory that the level of minimum wage affects inequality in gross-income distribution among employee-headed households. This is so whether one relates uniformly to the minimum income stipulated by agreement for the 1972–April 1987 period and the income enshrined subsequently in law, or whether one treats the minimum wage until the law was enacted as zero. This apparently reflects the low level of compliance with both the agreement (Yaniv, 1986) and the law (Flug and Kasir, 1993). This result may also trace to the lack of a stipulation in the Minimum Wage Law requiring consideration of household income. Theoretically, the Minimum Wage Law may reduce inequality among workers but not among households due to the structure of breadwinners in the household. This would occur even if compliance with the law were total and even if the law had a negligible effect on dismissals of subminimum-wage workers.
20
3. a.
Inequality and Macroeconomic Factors
Unemployment and Inequality
The unemployment rate climbed rather quickly, from 3.3 percent of the civilian labor force in 1979 to 8.9 percent in 1989, and rests at 8.7 percent today. It is difficult to state that this volatility reflects only employment changes that are typical of the business cycle, the type that naturally tends to disappear over time. For this reason, the relationship between unemployment and inequality has an importance that surpasses any question of smoothing business cycles in economic activity and employment. Several studies have shown a strong correspondence between the unemployment rate and the distribution of income among households headed by working employees in Israel.10 A rising unemployment rate has an expansionary effect on inequality and has been one of the main explanatory factors in Israel’s inequality trend in the past thirty years. In other words, when unemployment worsens, so does the relative situation of people in the lower income deciles.11
10
Dahan (1995), Achdut (1996), Silber and Zilberfarb (1994).
11
A similar result was obtained in the United States (Blinder and Esaki, 1978),
Great Britain (Nolan, 1988), the Philippines (Blejer and Guerrero,1990), and Sweden (Bjorklund, 1991).
21
No discussion of the effect of the unemployment rate on incomedistribution inequality among employees would be adequate without reference to the unemployed. The more unemployed there are, the more people there are whose incomes were zero before the Minimum Wage Law was enacted and remained paltry afterwards. An increase in the unemployment rate causes greater inequality if the unemployed earned low incomes during their term of employment. According to the CBS Manpower Survey, unemployment is more prevalent among the poorly schooled than among the well schooled. Additionally, the less schooling people have, the more harm an increase in the unemployment rate causes them. In other words, the upturn in unemployment rate exacerbated inequality not only among employed households but also among the population large.12 One of the main factors in the upturn in inequality in the late 1980s and early 1990s was the rising unemployment rate, which reflects the crowding out of part of the population—foremost members of disadvantaged groups—from the employment cycle. This large rather large pool of unschooled unemployed
12
Awwad (1998) found that unemployment has a regressive effect on
inequality in the population at large.
22
threatens to dislodge low-wage workers, allows their employers to pay them even less, and, thereby, exacerbates wage disparities.13 b.
Inflation and Inequality
Empirical research in Israel also shows a significant positive correlation between the inflation rate and the inequality index,14 indicating that the lower
13
Pressure on the wages of the disadvantaged also mounted due to the arrival
of foreign workers (beginning in late 1993), foremost in occupations that pay low-decile wages. It also has an erosive effect on the relative wages of workers in disadvantaged strata and, thereby, an expansionary effect on wage disparities. Israel’s population of foreign workers far exceeds that required to replace workers from the territories, whose numbers have fallen due to security problems. Today, one worker in ten in the business sector is a “guest worker,” a proportion higher than that of low-wage Israeli workers. The impact of this factor cannot be examined systematically due to data limitations. 14
Use of a shorter term than that used by Dahan (1995) and Silber and
Zilberfarb (1994) shows that inflation is not econometrically stable in its impact on inequality (Achdut, 1996; Awwad, 1998). Another noteworthy qualm is that the income-measurement method may create an artificial positive correlation between inequality and inflation. However, Dahan (1995) showed that the inequality-inflation nexus is preserved with hardly any change even when the measurement skew is taken into account.
23
deciles are unable to evade the relative erosion of their income. This finding is consistent with Artstein and Sussman (1991), who found that the standard deviation of wage changes rose in tandem with the escalation of inflation in 1984 and 1985 and contracted after the Economic Stabilization Program was applied. The finding also supports Sussman and Zakkai (1983), who showed that the wage structure in general government became less equal as the inflation rate rose. This is a thought-provoking finding in view of the wage accords that labor and management concluded during the period reviewed. For a long time, the agreement stipulated a wage ceiling beyond which a COLA would not be paid. Thus, in de facto terms, high wages were indexed at a lower rate than that declared (Leviathan, 1982). This is because the trade unions deliberately skewed the COLA agreement progressively to stanch inequality as the inflation rate rose. This progressive skewing, however, was not translated into practice, and inflation even served as a regressive tax. One way to explain this, perhaps, is to note that for low-wage earners the COLA agreement was a major factor in wage increases—under which circumstances only part of the real wage erosion was stanched—whereas wage changes among upper-decile earners were dictated more by personal contracts, which were better protected against price increases. These findings are important in an overall evaluation of fiscal and monetary policy that may affect both the unemployment rate and the inflation
24
rate. In the short term, of course, sustained high (low) inflation and low (high) unemployment for a short time may be substitutable. The discussion above allows us to add the aspect of income distribution to the set of considerations. To obtain a deeper understanding of this issue, research is needed on the direct effects of monetary policy on income distribution after interest payments (receipts). One presumes that individuals at the bottom of the income scale are net borrowers and that those at the top are mostly lenders. Changes in the Bank of Israel key rate may have at least a short-term effect on distribution of income after net interest payments. c.
Immigration and inequality
The mass immigration that began in late 1989 exacerbated the uptrend in income-distribution inequality in the first half of the 1990s (Dahan and BenPorat, 1997). The main transmission mechanism between immigration and income disparities was the escalation of unemployment to a rate surpassing that observed in 1966. This aggravated the erosion of the bargaining power of nonimmigrant Israeli workers in low income deciles. The flow of immigrants to unskilled jobs also took a toll on the bargaining power of low-wage nonimmigrants. Consequently, the relative wage of low-wage workers declined and economic disparities widened. Even though a high proportion of immigrants are in low deciles (as we discuss in the next section), the fraction of immigrants in the population was not found to have a significant effect on inequality in the 1990s. As stated, the
25
main impact occurred via the increase in the unemployment rate. The inconsistency between the findings of a latitudinal cross-section observation (at a given point in time) those based on a time series (a regression) is typical of many studies in labor economics. 4.
Israel from a Global Perspective
The World Bank, in a far-reaching project, has built a database of inequality that includes most countries around the globe.15 To assure a satisfactory level of uniformity and reliability, countries were included in the database only if their inequality data met several criteria. One of the criteria was full coverage of the population, and Israel was excluded from the sample on this account. The CBS Income Surveys omitted small localities that, before 1995, were home to about 20 percent of the country’s employee population. In 1995, the coverage rate climbed to 90 percent. Here, with the warranted degree of caution, we attempt to observe Israel’s inequality from an international perspective. Much has been written about the empirical correspondence between economic development and inequality, especially in the context of the Kuznetz Curve. The positive correspondence between inequality and per-capita GDP, as an indicator of economic development, stands out in Table 15.5. A glance at the data shows that no developed country whatsoever has severe inequality
15
Deininger and Squire (1996).
26
(the variance among developed countries is actually rather small) but that some developing countries have rather small inequality. This is the source of the alienator relationship that is termed the Kuznetz Curve.
Table 15.5. Inequality and Economic Development ($) No. of countries
GDP range
Avg. per-capita Gini index GDP 21 17,594-9,843 13,351 34.8 21 9,238-4,027 6,086 40.9 21 3,942-1,974 2,820 43.1 25 1,913-482 1,107 43.9 Israel 1992 9,843 9,843 33.9 Source: the Gini index is culled from Deininger and Squire (1996) and percapita GDP from Summers and Heston (1984). The figures for Israel pertain to net-income inequality among households.
International comparison gives us a yardstick with which we may determine Israel’s place. Israel rests at the bottom of the group of affluent countries (the first column in Table 15.5) and its level of inequality in net income distribution resembles that in those countries. In other words, Israel’s level of net income inequality is not exceptional relative to its state of economic development. In economic income inequality, however, as mentioned above, Israel’s level resembles that of South America. I did not find comparable figures on income inequality in additional countries. Israel is not unique in evincing an upturn in economic inequality. Economic research around the globe can help (to a limited degree) to identify additional factors that may be responsible for the increase in inequality in
27
Israel, since the Israeli economy is exposed to socioeconomic processes that resemble those sweeping the global economy. In the 1980s, inequality increased in most industrialized countries except for Germany and Italy (Gottschalk and Smeeding, 1997). Escalating inequality, especially in the United States, has attracted much attention that gave rise to documentation of the trends and attempts to explain them. The main factors bruited as explanations for the increase in inequality (mainly in the United States) are the increasing openness of industrial economies to flows of goods, services, and capital; a bias toward skills in economic growth; and the waning power of trade unions. These factors are interrelated and there is no consensus about the relative contribution of each to the increase in inequality. A recent study attempted to examine the effect of these factors on inequality in Israel.16 The lowering of import tariffs, a process that accelerated in 1987, made the domestic economy more receptive to inflows of goods and services. This openness advanced more vigorously, in relative terms, vis-a-vis countries less intensive than Israel in human capital. In the 1990s, the Israeli labor market
16
Klinov (1999) found no support for the premise that the increase in
inequality was occasioned by skill-oriented economic growth. However, she found indication of changes in the bargaining power of trade unions as a source of the rise in inequality.
28
also became increasingly exposed to the entry of foreign workers. The number of such workers (which is subject to doubt) surpasses those needed to replace Palestinian workers. In practice, the economy was opened asymmetrically, most of those who came were rather weak in human capital. The policy that led to this brought wage pressure to bear on people at the bottom of the domestic wage scale and may be partly responsible for the increase in inequality.
Table 15.6. Gini Index of Inequality among Households, U.S. and Israel, 1985–1997
1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997
Economic income Israel U.S.
Gross income Israel U.S.
Net income Israel U.S.
0.504
0.419
0.356
No survey
0.498 0.496 0.513 0.516 0.524 0.529 0.526 0.535 0.527 0.532 0.533
0.471 0.476 0.477 0.477 0.481 0.480 0.483 0.490 0.508 0.510 0.503 0.505 0.504
No survey
0.414 0.408 0.421 0.417 0.423 0.432 0.422 0.440 0.432 0.432 0.432
0.418 0.423 0.424 0.425 0.429 0.426 0.425 0.430 0.448 0.450 0.444 0.447 0.448
No survey
0.361 0.358 0.367 0.366 0.373 0.378 0.368 0.385 0.372 0.375 0.372
0.410 0.434 0.408 0.411 0.415 0.410 0.410 0.415 0.430 0.431 0.424 0.429 0.431
Sources: Israel—Table 15.1; U.S.—United States Bureau of the Census
The attenuation of trade unions’ power (the decline of the Histadrut) visa-vis business-sector employers is undoubted even though systematic
29
documentation of this trend is lacking.17 A trade union affects labor-market inequality in into main ways. First, like any social guild, a union wishes to widen wage disparities between organized workers and the unorganized and unemployed.18 Second, to unify the ranks, a union strives to attain for a uniform wage formula that corresponds to its members' personal characteristics. Naturally, the wage formula is based on overt and measurable traits such as years of schooling and job seniority. These conditions leave employers with little leeway in rewarding workers for the quality of their schooling and (innate) personal attributes. This explains why this factor helps
17
One cannot state with certainty how the power of trade unions in central
government has changed. Table 15.4 shows that the relative wages of the academically trained, a large proportion of whom are employed in general government, increased substantially in 1994–1996. This finding is consistent with the largesse of the general-government wage accords that were in effect during those years. Whether this largesse reflects an upturn in the union power of general-government labor is an open question. 18
Sussman and Zakkai (1997) identify another way in which a trade union
may aggravate inequality. These researchers document the fascinating story of the Israel Medical Association, which took action in the 1990s to widen wage disparities between nonimmigrant Israeli and newly landed immigrant physicians.
30
reduce (conditioned) inequality. Additionally, wage disparities may occur among members of the trade union in accordance with the relative strength of groups within the union. Labor at the Israel Electric Corp., Mekorot, and Bezeq can disrupt Israelis’ lives gravely. This ability is translated into a high relative wage with the union’s support. In return, the trade union acquires an important platform for struggle against employers to improve the wages of all members. From an empirical point of view, a strong trade union enhances the explanatory power of measurable traits such as schooling and age, and as a result, the unexplained remainder contracts. The long-term decline of the Histadrut has had a downward effect on inequality among “equals.” However, as the foregoing discussion showed, the contraction of a trade union's power does not necessarily reduce labor-market inequality. Another factor that has attracted some research attention is the winnertake-all phenomenon, described (although not for the first time) in Cook and Frank’s American best-seller (1995). According to this explanation, the upturn in inequality is rooted in technological and social changes that allow star performers in services (especially in entertainment) to command astronomical wages. An Israeli example, which drew impressive journalistic coverage, was the phenomenal increase in the wages of the television presenter Haim Yavin. More systematic evidence is the dramatic increase in the relative wages of executives, as we describe in the second part of this study. Admittedly, the
31
increase in inequality has been affected significantly by the ascent of the wages of individuals at the top of the pyramid, but the explanatory potency of that argument is unclear. In terms of similarity, Israel’s inequality trend has several unique factors. First, inequality in the United States began to increase in the late 1960s, whereas in Israel the process got under way in the late 1970s. Second, our survey of research on inequality in Israel shows that inflation, which climbed to three-digit levels in the 1980s, had an expansionary effect on incomedistribution inequality. This inequality contracted after inflation was checked in 1985 and rose again in the early 1990s. Third, the widening of inequality in the early 1990s coincided with mass immigrant absorption. No Western country has experienced such a process. The influx of foreign workers at that time may also have exacerbated economic disparities. The coincidence of the increase in inequality and immigrant absorption (along with the admission of foreign workers) and the findings of previous studies illuminate the importance of this factor in explaining the upturn in inequality. Fourth, Israel's unemployment rate has doubled since the late 1970s, and several studies have shown this to be a major a factor in the increase in inequality. Since unemployment did not increase in the United States, the upturn in inequality there cannot be traced to this factor. An additional factor, difficult to measure but evidently active in the
32
background, is the process of social atomization in Israel. It is difficult to state exactly when the process began, but by the late 1970s one could sense the change in Israel's social climate by observing its reflection in the change in the parliamentary balance of forces. In the 1980s, values such as cooperation, mutual assurance, and economic egalitarianism did not always pass the text in leading institutions such as kibbutzim, moshavim, and the Histadrut. The erosion of these values seems to have “rewritten” unwritten rules that had formerly imposed wage restraint. In the absence of these rules, executives at the top of the pyramid were able to secure mammoth salaries.
Part B: The Rise of Labor-Market Inequality
1.
Introduction
In the previous section, we saw that economic inequality among households has been rising in Israel since the late 1970s and that the distribution of labor income is central in this development. The widening of economic gaps coincided with important changes in the structure of the economy. In the mid1980s, the Israeli economy shifted from high inflation to a hyperinflation that the Economic Stabilization Program braked abruptly. The arrest of inflation forced firms to adjust to the new environment. The structural change that this entailed, coupled with other factors, caused the unemployment rate to rise by the end of the 1980s to a level twice that of the late 1970s. In the 1990s, the Israeli economy had to undergo further restructuring to take in hundreds of
33
thousands of immigrants over a rather short period of time. This work explores the origins of the increase in labor-market inequality between 1980 and 1996. It focuses on three factors that may be responsible for the widening of labor-wage disparities. The first is a change in the structure of return on human capital. By examining the impact of this factor, we may determine whether wage inequality increased because of a relative upturn in the return on higher schooling. The second possible factor is a change in the distribution of human capital, i.e., wage disparities may have widened because of an increase in education disparities. Obviously, these two factors are not independent of each other. The third possible culprit is a change in inequality among “equals,” i.e., wage disparities among workers who have identical measurable traits, where these disparities are not explained by inequalities in return on various dimensions of human capital or by disparities in schooling. Such an analysis allows us to discuss economic inequality in terms of prices (return on human capital) and quantities (ownership of human capital) that are less evasive in terms of systematic economic analysis. One can identify the channels along which economic events, such as disinflation and mass immigration, affect inequality in the labor market. For example, one may examine whether the shift to hyperinflation or the steep increase in labor supply affected economic inequality via prices, quantities, inequality among “equals,” or any combination of these.
34
In making income-redistribution policy, it is important to pinpoint the origins of the increase in labor-market inequality. Government can act in several ways to influence the degree of inequality among households. It can tackle education disparities at the micro level and, by so doing, have a direct effect on the dispersion of human-capital ownership and an indirect effect on the structure of return on human capital. It can intervene after the fact (i.e., after the distribution of economic income has been determined) by setting rates and types of taxes and levels of transfer payments. Here, too, it is important to identify the sources of inequality to choose the optimum intervention tools. 2.
The Data
This work is based on the Central Bureau of Statistics Manpower and Income Surveys for 1980–1996. Each year, the CBS conducts four manpower surveys, in which one-fourth of respondents are also asked about their income. The figures on these one-fourth of respondents are published each year as part of the Household Income Survey. These surveys provide information about 80 percent of Israel's employee population, on the basis of a sample of 6,000 households that contain about 20,000 individuals. The rate of coverage increased in 1995, especially in respect to the proportion of non-Jews in the sample. The information includes a comprehensive itemization of sources of income—among which labor income is central—and particulars about individuals' traits such as schooling, age, year of immigration, nationality, and labor patterns. The work focuses on the wages of full-time employed men, the
35
number of whom in the 1996 survey was approximately 2,800. (Their numbers range from 2,000 to 3,000 during the period reviewed.) The next section explains why this population group was chosen. The information obtained from one source is divided into two surveys: of manpower and of income. The manpower survey includes detailed information on various aspects of labor patterns but excludes information on income. The income survey includes detailed information on income but confines information on labor patterns to a limited appendix. The main customer for the income survey is the National Insurance Institute (NII). The NII uses the survey mainly to calculate two indicators that have become part of the landscape of economic data: the poverty line and the extent of poverty. The CBS produces a collection of data that is tailored to the needs of the NII, and this collection is used by other institutions such as the Bank of Israel and academic institutes. The income survey data may be used to estimate wage regressions. Such estimations, however, may be limited because the Income Survey is tailored to the needs of the NII, which are not always consistent with the needs of researchers. For example, years of schooling are presented in aggregate form for some of the survey period. Since one cannot know exactly how many years of schooling workers have, researchers cannot estimate the return on human capital in the accepted fashion. The CBS does possess the full information and is always willing to disclose it to all comers. For this purpose, it is necessary to
36
unify the data in the two surveys (Manpower and Income). The large budget and lengthy wait that are needed to unify the surveys for each of the years 1980-1996 impede research on economic inequality in Israel. In this work, I used aggregate years of schooling even in years for which detailed information on this indicator is available. I accepted this compromise to maintain uniformity of the variable that expresses one of the most important dimensions of human capital. This uniformity allows us to compare the regression coefficients over time. The information on the last type of school that the worker attended, a possible indicator of schooling quality, is available to researchers for some years only and, for this reason, its long-term trend cannot be monitored. These compromises (and other compromises in other studies) evoke several puzzling questions: why does the CBS split the information into two surveys? Why is there no quarterly information on household income? Is the ten-month lag in presenting the income-survey data necessary? 3.
The Rise of Labor-Market Inequality
This part of the study focuses on inequality in the labor market. The research population in this study is composed of employed full-time men aged 25–65. The variance of the natural logarithm of wage per hour worked in this population serves in this study as an indicator of labor-market inequality. Inequality in wage per hour worked of full-time employed men reflects an important sector of the labor market but not the entire employed population.
37
The population chosen for the study reflects a tradeoff between broad coverage and limited reliability of data and incomplete coverage and high reliability. The labor-force participation rate of women increased significantly during the period reviewed. Men were chosen in order to monitor changes in inequality that are not related to the changes in women’s working habits. Full-time employed were chosen because the wage per hour worked of part-time employees is susceptible to relatively large measurement errors and severe volatility. Even if one accepts the considerations that informed the choice of population, one may ask (justifiably) about the degree of correspondence between inequality in wage per hour worked of full-time employed men and accepted indicators of inequality in household income. Table 15.7 shows correlation coefficients for variance in the natural logarithm of wage per hour worked among full-timed employed men and Gini indices for economic income, gross income, and net income of employee-headed families. The table shows a rather strong correlation between the indicator that this study uses to express labor-market inequality (variance) and the inequality indices (Gini coefficients) for the different definitions of income. In 1980–1996, there was a rather strong correlation (roughly 0.8) between the variance of (the natural logarithm) labor wage and the Gini index for the distribution of household economic income. This finding alludes to the existence of a strong correspondence between a man’s income and that of his spouse, and with
38
capital income. Consequently, inequality among men is transmitted and maintained, to a considerable extent, among households as well. Figure 15.2 illustrates the inequality in wage distribution per hour worked among full-time employed men in 1980-1996. Labor-market inequality increased steadily between 1980 and 1985, as the economy shifted from high inflation to hyperinflation. Labor-market inequality contracted after the successful Economic Stabilization Program was applied. Inflation settled at approximately 20 percent, but concurrently the unemployment rate climbed to about 9 percent in 1989, almost twice the rate in 1980. Wage disparities widened again in 1990 and subsequent years, concurrent with the absorption of mass integration. Inequality narrowed in 1996 but remained much higher than the 1980 level. What are the factors to which the increase in labor-market inequality between 1980 and 1996 should be traced?
39
Table 15.7. Correlation among Indicators of Inequality: 1980-1996a Gini index of economic income 1
Gini index of gross income —
Gini index of net income —
Variance of wage per hour worked —
Gini index of economic income (before transfers to employee-headed households) Gini index of 0.82 1 — — gross income (before transfers (5.27) to employeeheaded households) 0.77 0.85 1 — Gini index of net income (after transfers to and (4.56) (6.06) tax payments by employee-headed households) Variance of wage 0.69 0.78 0.44 1 per hour worked (3.69) (4.7) (2.09) among full-time employed working-age men a. The t values are shown in parentheses under the partial correlation coefficients. Insert Figure 15.2
.4
Origins of the Increase in Labor-Market Inequality
We see that the upturn in labor-market inequality was translated into a similar trend of inequality in economic-income distribution among households. In this part of the study, we present a method that allows us to identify the factors
40
responsible for the increase in inequality in the labor market.19 The method points to four sources of inequality: (a) inequality resulting from inequality in ownership of characteristic quantities of human capital, such as schooling and occupational experience; (b) inequality stemming from differences in the return on human capital; (c) the product of these two factors; and (d) inequality that the observable variables cannot explain. The central feature of this method is a conventional regression of the human-capital equation (Mincer’s equation), in view of which the natural logarithm of wage per hour worked is explained by variables that represent different aspects of the level of a worker’s human capital. The change in variance of the explained variable (wage per hour worked) may flow from changes in this variance that reflect the distribution of ownership of human capital, changes in the regression coefficients that express the market price of the human capital, and changes in the variance of the unexplained remainder that represents inequality among “equals.” A formal phrasing of the human-capital equation follows:
19
Juhn, Murphy, and Pierce (1993) used a similar method to examine the
origins of the increase in U.S. inequality in 1963–1989. Other indicators of inequality, such as Gini and Theil, can be reduced to components but do not show clearly (intuitively) the distinction between prices and quantities as does the method presented above.
41
(1)
yi,t = Xi,tβt + εi,t
where yi,t is the wage of individual i in year t; Xi,t is a vector of several observable variables that represent different indicators of human capital, such as schooling and job seniority; and βt,t is a vector of regression coefficients for the observable variables, by means of which the value of ownership of physical quantities of human capital may be translated into financial income. εi expresses the disparity between the actual income of individual i and his income as measured for an assumed degree of ownership of human capital and the observed prices of human capital. Below we write an identical equation for period t-l: 2.
yi,t–1 = Xi,t–1βt–1 + εi,t–1 We use the two equations to calculate the long-term change in wage
variance (in log terms) after several simple manipulations: 3.
V(yi,t) – V(yi,t–1) = A + B – C + D Where: A = [V(Xi,t)-V (Xi,t-l)]βt2 B = V(Xi,t)(βt2–Bt-l2) C = [V(Xi,t)-V(Xi,t-l)] (βt2–Bt-l2) D = V(εi,t)-V(εi,t-l) Equation (3) includes the four components that make up the variance in
wage per hour worked (in log terms), the variance that represents labor-market inequality in this study. The first factor, A, expresses the change in the
42
distribution of quantities of the explained variable (human capital), which is composed of the sum of changes in variance of the various indicators of human capital and the correlation among them. The second factor, B, reflects the effect of long-term changes in prices of human capital on inequality in wage per hour worked for a distribution of quantities of human capital in the current period (which corresponds to a price index according to Laspeyres20 Equation (3) may also be presented in terms of quantities of the previous period which corresponds to a price index according to Paasch or any combination thereof. The third factor, C, is the product of the change in both of the two first-mentioned factors; it is expected to have a small quantitative effect on inequality. The fourth factor, D, reflects the effect of changes in variance of the unexplained remainder on wage inequality. These are the wage-income disparities that are not explained by differences in years of schooling, job seniority, or other observable traits included in the regression. In this study, we use the term “inequality among ‘equals’” to represent these unexplained wage disparities. Since this term represents a linguistic “short cut” only, we place the word “equals” in quotation marks. “Equals” expresses quality in measured
20
Inconsideration of the degree of significance of the regression coefficients is
one of the drawbacks of this method.
43
traits. Obviously, other traits may that affect wage level even if we lack information on them.
.5
Results
A human-capital equation was estimated for each of the years 1980–1996, including potential experience, potential experience squared, dummy variables for five categories of reported years of schooling,21 dummy variables for seven groups of occupations,22 a dummy variable for household situation (married is
21
The six schooling cohorts are 0–4, 5–8, 9–10, 11–12, 13–15, and 16+. The
0–4 cohort was omitted from the regression. The Income Survey for 1996 makes it possible to compare the explanatory power of the total regression on the basis of an exact number of years of schooling as against schooling cohorts as an explained variable. It was found that the availability of precise data on years of schooling improves the explanatory power of a regression only slightly. 22
The eight occupation groups are academic worker; liberal and technical
professionals; managers; white-collar workers; agents, salespeople, and service workers; skilled farm workers; skilled industrial and construction workers; and unskilled workers. The group of unskilled workers was omitted from the regression. A more refined division of occupations (sixteen groups) elicits similar results and most occupation dummy variables are not significant at an accepted level.
44
assigned the value of 1; others are 0), and a dummy variable for nationality (a Jew is assigned the value of 1; others are 0). These regressions provide a time series of regression coefficients and annual variance of the unexplained remainder; these data are needed to analyze the four components of inequality. The regression omitted the main industry groups in which the workers are employed in order to avoid confusion between changes originating in labor demand and changes in labor supply.23 Averages of the explanatory variables appear in Table 15.8. Table 15.9 presents the results of the regression for the entire period investigated (1980–1996). The table shows that a one-year increase in potential experience (for an individual with zero experience) leads to an annual wage increase of 4.1 percent (on average over the entire period). Over the course of the period, the market price of a year of seniority is quite volatile. The return on a year’s experience climbed from 3.6 percent in 1980 to a peak of 5.5 percent in 1990 and slipped to 3.2 percent in 1996. In contrast, the trend in the return on schooling is clear, albeit with rather severe year-on-year fluctuations. Table 15.9 presents the regression coefficients of five schooling categories; the value of the coefficients of each schooling group reflects an annual percent wage increase relative to the wage
23
Eckstein and Weiss (1998a,b), who looked into the immigrants’ labor
supply, use the same methodology.
45
of workers with 0–4 years of schooling. In 1996, for example, a worker with 16 or more years of schooling earned 60 percent more than a worker with 0–4 years of schooling but the same traits (among the traits included in the regression). Sixteen or more years of schooling provided a smaller return in 1995 and 1996 than in the early 1980s. However, the added increment of highly schooled workers who had just reach the country at that time may have skewed the return on higher schooling significantly (as we discussed below.).
Table 15.8: Mean Values of Explanatory Variables
46
Potential experience (years) Married (%)
1980–1981 Population at large 29.6
1995–1996 Nonimmigrants 28.4
1995–1996 Population at large 28.3
91.8
84.7
85.1
96.4
83.0
84.9
0–4
6.6
1.8
1.9
5–8
24.6
12.0
10.5
9–10
14.9
12.4
12.5
11–12
26.8
36.8
33.8
13–15
12.0
17.3
20.2
16+
15.1
19.6
21.1
Academic
9.5
11.5
11.7
Liberal
9.8
8.7
8.6
Managerial
6.6
8.9
7.6
White-collar
17.0
11.1
9.8
Sales
9.8
12.6
11.4
Agricultural
1.6
1.3
1.3
Industrial
38.4
39.2
41.4
Unskilled
7.3
6.8
8.2
0
0
Jewish (%) Years of schooling (%)
Occupation (%)
Immigrants (%)
Table 15.9. Wage Regressions, 1980–1996a
17.1
47
1980 3.246
1981 1.707
1982 2.749
1983 1.481
1984 5.607
1985 -0.319
1987 0.352
0.036
0.036
0.036
0.036
0.031
0.049
0.051
-0.0005
-0.0005
-0.0005
-0.0005
-0.0004
-0.0007
-0.0007
0.230
0.292
0.193
0.170
0.087
0.158
0.114
0.207
0.244
0.121
0.212
0.331
0.251
0.177
5–8
0.213
0.113
0.141
0.145
0.043
0.118
0.207
9–10
0.363
0.261
0.247
0.271
0.196
0.242
0.372
11–12
0.467
0.407
0.388
0.454
0.339
0.366
0.532
13–15
0.570
0.504
0.507
0.613
0.459
0.544
0.722
16+
0.625
0.616
0.580
0.655
0.552
0.627
0.800
Academic
0.245
0.271
0.326
0.207
0.348
0.339
0.157
Liberal
0.171
0.300
0.244
0.198
0.209
0.333
0.150
Managerial
0.260
0.333
0.350
0.332
0.423
0.484
0.227
White-collar
0.055
0.157
0.104
0.036
0.089
0.204
0.043
-0.127
-0.035
-0.128
-0.101
0.002
0.031
-0.154
Agricultural
0.015
-0.112
0.014
-0.227
-0.097
-0.081
-0.153
Industrial
0.021
0.107
0.097
0.068
0.101
0.158
0.057
Total variance
0.255
0.303
0.277
0.321
0.345
0.327
0.316
:Unexplained
0.182
0.215
0.201
0.235
0.268
0.232
0.218
Constant Potential experience Experience squared Married Jewish Yrs of schooling
Occupation
Sales
No. of observations
2,357
2,301
2,266
2,105
2,101
2,153
1,942
48
1988
1990
1991
1992
1993
1994
1995
1996
0.767
1.073
0.959
1.227
1.316
1.544
1.427
1.640
1.754
0.045
0.044
0.055
0.045
0.045
0.038
0.038
0.037
0.032
-0.0006
-0.0006
-0.0007
-0.0006
-0.0006
-0.0005
-0.0005
-0.0004
-0.0004
0.116
0.120
0.105
0.150
0.094
0.078
0.156
0.148
0.109
0.144
0.104
0.104
0.014
0.103
0.086
0.091
0.090
0.106
0.112
0.062
0.191
0.193
0.137
0.060
0.148
0.002
0.280
0.294
0.175
0.309
0.242
0.196
0.102
0.236
0.143
0.276
0.375
0.314
0.413
0.365
0.364
0.234
0.386
0.247
0.419
0.520
0.463
0.595
0.463
0.340
0.223
0.364
0.229
0.458
0.593
0.610
0.673
0.543
0.542
0.299
0.505
0.359
0.592
0.291
0.197
0.184
0.417
0.391
0.608
0.637
0.782
0.631
0.232
0.167
0.135
0.327
0.458
0.516
0.591
0.542
0.465
0.313
0.373
0.276
0.536
0.588
0.729
0.864
0.826
0.794
0.113
0.028
-0.019
0.215
0.267
0.414
0.410
0.457
0.413
-0.072
-0.068
-0.086
-0.026
-0.007
0.134
0.234
0.351
0.241
-0.114
-0.193
-0.199
0.056
-0.129
-0.134
0.024
0.088
-0.040
0.086
0.039
0.021
0.083
0.110
0.160
0.215
0.224
0.165
0.274
0.301
0.284
0.306
0.355
0.331
0.362
0.372
0.338
0.194
0.207
0.204
0.215
0.255
0.242
0.251
0.250
0.226
1,983
a.
1989
2,136
2,205
2,427
2,325
2,101
2,335
2,776
Numbers in bold denote 1% significance; underlined numbers denote 5% significance.
2,816
49
Seven occupational groups were included in the regression. The occupation of “unskilled worker” was omitted and each coefficient expresses an annual percent wage increase in the occupations that appear in the regression in the equation for unskilled workers. The results of the regression show, predictably, that managerial workers occupy the top of the wage scale. The relative wages of academic and managerial workers rose meteorically between the early 1980s and the 1990s. The return on marriage sank from a relatively high level of more than 20 percent in the early 1980s to 10 percent in the mid-1980s and remained on that plane for most of the period. This shows that marriage is not what it used to be. . . . On average for the years 1980–1996, Jewish workers earned 14 percent more than non-Jewish workers.24 This increment was greater in most of the
24
This finding may be consistent with discrimination against Arab workers,
but the wage disparity between Jewish and non-Jewish workers who have the same observable traits does not necessarily trace to discrimination and racism. This is because the regression does not include all characteristics of workers, such as quality of schooling; especially, it does not take account of possible differences in the attributes and location of the worker’s job. This doubt comes up because it is difficult to presume the existence of discrimination against singles, widowers, and divorcees as opposed to married persons, even though a
50
1980s and declined to 10 percent in the first half of the 1990s. Furthermore, in the 1990s, this wage increment was not always significant to the accepted extent. Table 15.10 (the column farthest to the right) shows the estimated effect of these factors on labor-market inequality. The impact of the changes on the structure of human-capital return are obtained by multiplying the prices—the estimated annual regression coefficients for each of the human-capital traits that were included in the regression—by the inequality in quantities (i.e., the ownership of these traits) in 1996. In other words, changes in this component may originate only in changes in the regression coefficients, commensurate with their relative importance.
Table 15.10. Factors in the Increase in Labor-Market Inequality (1990–1996 mean vs. 1980)
Change in inequality Unexplained portion Structure of human-capital
Total population without dummy variable for immigrants 8.1 5.6 3.6
Total population with dummy variable for immigrants 8.1 2.1 **
Nonimmigrants only
6.9 2.8 3.4
married man earns more than an unmarried man with identical observable traits.
51
return Ownership of human capital Ownership and structure of return
0.4 2.8
** 6.0
0.9 4.1
A similar computation was performed to estimate the effect on wage inequality of inequality in ownership of quantities of various human-capital traits. Inequality in the various indicators of human capital, such as years of schooling and occupational seniority (and the variance shared by all indicators), measured for each year, was multiplied by the prices of the indicators as estimated for 1996. On average, labor-wage inequality was 0.336 in 1990–1996 as against 0.255 in 1980.25 About 65 percent of the upturn in wage inequality among fulltime employed men stems from an increase in inequality among “equals,” which, as stated, was measured in accordance with variance in the unexplained
25
A wage equation was also estimated with use made of a sample that unifies
all observations in 1990–1996. Here it was found that the main regression coefficients do not swerve substantially from the average of the coefficients estimated for each year. The small number of years illustrates with greater emphasis the need to present the data for each year separately. In view of the negligible differences in the size of the coefficients, I preferred to maintain consistency in presenting the single-year data rather to gain the econometric advantage of unifying the observations.
52
remainder. The change in prices (i.e., the structure of return on human capital) explains almost the rest of the increase in inequality. The growth of inequality in quantities (i.e., changes in inequality in ownership of indicators of human capital) had a negligible impact on labor-market inequality (Table 15.10). The specific contribution of changes in return on human capital (prices) as against changes in inequality in human-capital ownership (quantities) are only slightly sensitive to the choice of base year.26 There were no significant year-on-year fluctuations in the effect of quantity inequality and structural changes in human-capital return on wage inequality. Predictably, price volatility is greater than quantity volatility. Since owning various indicators of human capital is somewhat like owning an inventory, one expects inequality in human-capital quantities to shift slowly. Prices, however, may change sharply. In 1995, there was a relatively steep relative increase (although not an absolute increase) in the contribution of quantity inequality. This occurred because the coverage rate of the Arab
26
Notice that according to the method invoked here to take apart the
inequation, the datum for each year is compared with the base year chosen (and not with the year preceding it).
53
population increased in the income survey for 1995, as reflected in a significant upturn in variance of the nationality dummy variable.27 Figure 15.3 illustrates the contribution of inequality among “equals” to wage inequality, as measured in view of the variance in the unexplained remainder in the human-capital regression. The upper curve in the figure almost perfectly mirrors the wage-inequality trend shown in Figure 15.2. Inequality among “equals” increased when high inflation moved into hyperinflation in the mid-1980s, contracted immediately after inflation was checked, and rose again with the onset of mass immigration in the 1990s. The spot estimate for a regression coefficient is single-variable; the variance of the remainder appears on the right-hand side, and the wage variance is positioned on the left-hand side and is roughly equal to 1. Insert Figure 15.3 The correlation between inflation and inequality corresponds to previous studies surveyed in Part A of this work. The findings here are novel in that they show the trajectory along which the effects of inflation are transmitted.
27
The variance of the dummy variable (which we mark with a V) that receives
the value of 1 or 0 equals V=a(I-a)2+a2(1–a), where a is the mean of the dummy value. This mean is equal to the proportion of observations that receives the value 1. It may readily be seen that the variance of the dummy variable is a negative function of a if a is greater than 0.5.
54
The results here suggest that inflation acts mainly to increase the component of randomness (luck) in wage and has a negligible effect on the return on schooling. In the early 1990s, the narrowing of wage disparities after the Economic Stabilization Program was halted and labor-market inequality began to expand again. The coincidence of the widening of two forms of inequality—laborwage and unexplained inequality—and the onset of mass immigration suggests that the two are correlated. The next section of the study investigates how the massive influx of labor in the 1990s affected labor-market inequality. 6.
The Effect of Mass Immigration28
The indicators of immigrants' human capital are different from those of nonimmigrants. This may attenuate the predictive power of observable variables such as years of schooling and potential experience, which are reflected in an upturn in inequality among “equals.” Practically speaking, human capital is measured in this study, as in many other studies, on the basis of reported years of schooling and potential occupational seniority.
28
Dahan and Ben-Porat (1997) found that immigration plays an important role
in the upturn in household-income inequality. For a thorough analysis of the intake of immigrants in the labor market, see Eckstein and Weiss (1998) (1998a,b).
55
The use of years of schooling fails to take account of disparities among workers in quality of schooling; consequently, the quality of the wage disparity prediction is impaired. The quality of schooling acquired outside of Israel is different from that acquired by the Israel-born in the Israeli education system.29 The loss of schooling (those elements of schooling that are specific to the Soviet economy) upon immigration to Israel, the lack of fluency in Hebrew, and other acculturation difficulties make it hard for immigrants to fulfill the earning potential that the reported schooling ostensibly creates. The transition from the Former Soviet Union to Israel also eroded the human capital that is reflected in occupational seniority. The Israeli market priced the human-capital indicators differently for a significant added increment of workers than for nonimmigrant workers, and this may explain the increase in the 1990s in the unexplained portion of labor-market inequality. Table 15.11 presents regressions for 1990–1996 that make special reference to immigrants (workers who reached Israel after 1989). The estimated wage equation expresses the (possible) different effect of years of schooling and years of (potential) occupational seniority on the wages of
29
Several studies have estimated the human capital loss occasioned by
migration: Friedberg (1992), Flug, Kasir, and Ofer (1992), Eckstein and Shachar (1995), Beenstock and Ben-Menachem (1995), Weiss and Gotlibovski (1995).
56
immigrants, in accordance with variables of interaction between these variables and the immigrants’ dummy variables. It is clear that the market prices the immigrants’ occupational seniority differently. The coefficient of the variable of interaction between occupational seniority and an immigrant (AGE_IM) reflects a reality in which, in the 1990s, an additional year of seniority (in the first year) gives a nonimmigrant Israeli worker a 5 percent increase and an immigrant worker only 3 percent. This coefficient is significant at the accepted level throughout the period of immigration intake. In contrast, there is no unequivocal indication that immigrants’ schooling is less rewarded than nonimmigrant workers’ schooling. The interaction coefficients between schooling and the dummy variable for immigrants is not significant in most of the 1990s. However, in years when these coefficients were significant they indicate a substantial negative effect on the immigrants' wages. Moreover, the dummy variable for immigration (with no interaction whatsoever) is not significant in most immigrant-intake years after the special pricing of immigrants' schooling and occupational seniority is taken into account. Strangely, this coefficient was positive and significant in 1994. This result underscores the importance of explicit representation of the trajectory that makes immigrants' wages different from those of nonimmigrants. Table 15.11. Wage Regressions, 1990–1996a
57
1990 0.942
1991 1.212
1992 1.243
1993 1.347
1994 1.286
1995 1.553
1996 1.651
0.055
0.048
0.050
0.052
0.045
0.048
0.045
-0.0007
-0.0006
-0.0006
-0.0007
-0.0005
-0.0006
-0.0005
0.109
0.159
0.086
0.072
0.149
0.139
0.092
0.107
0.028
0.121
0.142
0.113
0.133
0.149
5–8
0.190
0.188
0.177
0.076
0.211
-0.060
0.200
9–10
0.316
0.279
0.303
0.199
0.363
0.137
0.259
11–12
0.416
0.404
0.458
0.318
0.531
0.254
0.422
13–15
0.612
0.617
0.613
0.523
0.695
0.386
0.625
16+
0.674
0.657
0.808
0.666
0.815
0.523
0.776
Academic
0.205
0.302
0.172
0.302
0.434
0.610
0.404
Liberal
0.133
0.195
0.235
0.225
0.362
0.376
0.265
Managerial
0.276
0.373
0.339
0.381
0.573
0.584
0.526
Constant Potential experience Experience squared Married Jewish Yrs of schooling
Occupation
White-collar
-0.018
0.094
0.072
0.184
0.194
0.288
0.220
Sales
-0.080
-0.102
-0.097
-0.002
0.111
0.224
0.095
Agricultural
-0.197
-0.048
-0.196
-0.083
-0.038
0.028
-0.089
0.030
0.030
0.058
0.082
0.140
0.175
0.104
n.a.
0.397
-0.125
0.306
-0.237
0.154
0.264
Immig.*9–10
-1.400
0.060
-0.438
0.097
-0.448
0.022
0.203
Immig*11–12
-1.389
0.151
-0.364
0.090
-0.501
-0.044
0.076
Immig*13–15
-2.196
-0.077
-0.569
-0.123
-0.674
-0.123
-0.073
Immig.*16+
-1.467
-0.019
-0.733
-0.308
-0.690
-0.137
-0.091
Immig.*exp.
-0.030
-0.015
-0.017
-0.020
-0.014
-0.019
-0.020
Immigrant
2.174
-0.037
0.610
0.294
0.623
0.394
0.306
Total variance
0.284
0.306
0.355
0.331
0.362
0.372
0.338
Unexplained
0.200
0.189
0.216
0.193
0.211
0.220
0.189
Industrial Immigrant*5–8
No. of observations
a.
2,205
2,427
2,325
2,101
2,335
2,776
2,816
Numbers in bold denote 1% significance; underlined numbers denote 5% significance.
Figure 15.3 sketches the trajectory of the unexplained component of inequality according to two different formulations of the wage equation: with
58
and without special reference to immigrants. The variance of the remainders in the wage equation in Table 15.9 were compared with the corresponding figure elicited by the wage equation in Table 15.11. The figure points clearly to the intake of immigrants in the labor market as the cause of the increase in the unexplained fraction of inequality in the 1990s. The analytical framework set forth above does not allow us to distinguish between prices and quantities because the formulation of the wage equation does not maintain a constant form over time. As Table 15.11 shows, the wage equation for 1990–1996 includes variables that, for obvious reasons, were not included in the 1980–1989 wage equation in Table 15.9. In other words, in the transition from 1989 to 1990, a variable that affected both quantities and prices came into play. It is possible to isolate the effects of prices (or quantities) only when one can hold one of the factors constant. Furthermore, the significance of several variables in Table 15.11 is quite volatile. These sharp fluctuations make it impossible to apply a uniform equation over time because the methodology of this study assigns an insignificant coefficient the value of 0. Insert Figure 15.4 Much of the increase in labor-market inequality in the 1990s traces to changes in the distribution of ownership of human capital and the structure of return on human capital (Table 15.10). However, as stated, the specific contribution of each of these factors cannot be estimated. The increase in the unexplained portion also played a role in the upturn in inequality in the 1990s
59
relative to the early 1980s. However, the effect of this factor is smaller than that estimated on the basis of a wage equation that does not take account of the immigrants’ uniqueness. This analysis leaves several important questions unanswered. For example, did the labor-market inequality increase because nonimmigrants obtained a higher return on higher schooling? Did the increase in inequality originate in a widening of disparities in nonimmigrants' schooling? Did the inequality widen because immigrants congregated at the bottom of the wage scale? Did it expand due to relatively severe inequality among the immigrants themselves? To answer some of these questions, a wage equation was estimated for a sample composed solely of full-time employed nonimmigrants. As before, we estimated wage inequality among nonimmigrants by measuring the variance of the natural logarithm of wage. Figure 15.4 shows that the increase in inequality documented for the entire working population is also typical of nonimmigrant working Israelis. By excluding immigrants from the sample, we may investigate the trend in inequality among nonimmigrants without having to rephrase the wage equation. Of course, inequality among nonimmigrants is not independent of the impact of immigrant intake in the labor market. Table 15.12 presents the results of the regressions for 1990–1996. The effects of the changes in the
60
distribution of human-capital ownership, return on human capital, and the unexplained inequality were computed on the basis of these regressions.
Table 15.12. Wage Regressions, 1990–1996a (immigrants excluded from sample) 1990 0.929
1991 1.189
1992 1.215
1993 1.288
1994 1.236
1995 1.504
1996 1.618
0.056
0.051
0.053
0.056
0.048
0.051
0.050
-0.0007
-0.0006
-0.0007
-0.0007
-0.0006
-0.0006
-0.0006
0.107
0.154
0.097
0.088
0.148
0.143
0.062
0.107
0.031
0.103
0.121
0.113
0.143
0.144
5–8
0.189
0.184
0.174
0.069
0.207
-0.077
0.194
9–10
0.315
0.275
0.304
0.193
0.356
0.116
0.255
11–12
0.415
0.403
0.462
0.317
0.529
0.236
0.421
13–15
0.612
0.622
0.620
0.530
0.707
0.383
0.630
16+
0.679
0.678
0.829
0.674
0.834
0.542
0.790
Academic
0.198
0.243
0.135
0.316
0.415
0.552
0.370
Liberal
0.138
0.165
0.229
0.234
0.343
0.350
0.250
Managerial
0.277
0.344
0.327
0.388
0.568
0.571
0.502
White-collar
-0.014
0.075
0.067
0.212
0.201
0.293
0.205
Sales
-0.070
-0.124
-0.088
0.029
0.134
0.234
0.096
Agricultural
-0.192
-0.065
-0.142
-0.082
-0.116
0.021
-0.047
0.033
0.015
0.051
0.112
0.164
0.189
0.092
0.279
0.294
0.337
0.316
0.348
0.366
0.329
0.197
0.193
0.227
0.227
0.220
0.227
0.197
Constant Potential experience Experience squared Married Jewish Yrs of schooling
Occupation
Industrial Total variance Unexplained No. of observations
a.
2,185
2,245
2,050
1,758
1,966
2,303
Numbers in bold denote 1% significance; underlined numbers denote 5% significance.
2,331
61
Table 15.10 summarizes the specific effects of the three causes of increase in inequality among working nonimmigrants. It shows that the changes in the structure of return on human capital were central in the increase in inequality among working nonimmigrants. Figure 15.5 points to a substantial increase, starting with the advent of mass immigration, in the specific contribution of price changes to the upturn in inequality. The significant increase in the relative wages of holders of managerial and academic posts was especially striking. Managers earned 42.5 percent more than unskilled workers with the same observable traits (e.g., occupational seniority) in 1990–1996 against 26 percent in 1980. The return on schooling for persons with thirteen or more years of study increased in 1990s relative to the early 1980s and the return for those with twelve years of study or less declined. Changes in unexplained inequality also played a role in the increase in inequality in the 1990s among working nonimmigrant Israelis. In contrast, changes in the distribution of ownership of human capital had a negligible impact. Insert Figure 15.5. 7.
Inequality among “Equals”
The human-capital equation estimated in this study explains only about onethird of the wage disparities in the labor market. This is due to limitations in our information about workers' traits, places of work, and occupational
62
attributes. In a certain sense, our inquiry is reminiscent of the man who lost his key under a streetlight (although the search should begin under the light). One may point to several factors about which information is limited (or wholly absent) that may create wage disparities among workers with identical reported traits (such as schooling and occupational experience): measuring problems, differences in innate aptitudes, racism and discrimination, and differences in the quality of the human capital. Most of the aforementioned factors are responsible for the large extent of inequality among “equals” at a given point in time, but one cannot know which of them was the main culprit in the upturn in this aspect of inequality. The prime candidate to explain this finding, it seems, is a possible widening of disparities in schooling quality. We possess raw information only on workers’ actual level of schooling and occupational experience.30 Years of schooling
30
An indirect measurement of occupational seniority as performed in this
study (number of years of seniority of a Jewish worker is equal to the workers age less the total years of study less nine additional years) is different from actual seniority, on which no information is available. Even this measurement masks disparities in human-capital quality that may be responsible for inequality among “equals.” For example, more frequent movement from job to job (shortening of career phenomenon), which involves a loss of human capital, may affect the trend in inequality among “equals.”
63
and potential years of experience mask possible disparities in the quality of the human capital. As we have seen, immigrants provide a marked example of this. The haredim (“ultra-Orthodox”) are another population group for which it is difficult to predict labor wage on the basis of years of schooling. The schooling delivered by haredi education institutions is intrinsically different from that acquired by other Israelis in the public and academic education system.31 Substantial disparities in schooling quality may be discovered among nonharedi nonimmigrant Israelis as well. The quality of human capital of workers who report having completed twelve years of study and received matriculation certificates does not resemble that of someone who completed twelve years of study in a vocational school and did not earned a matriculation certificate. The same may be said about the quality of schooling of teachers' college graduates as against university graduates. The increase in inequality among “equals” is consistent with disparities in the quality of schooling delivered by the
31
Dahan (1998), Berman and Klinov (1997), and Berman (1998) noted the
differences between haredi and general schooling. The information in my possession does not allow me to refer specially to the haredi population in each year at issue. I should note that in 1996 (for which I have information about the haredi population) fewer than 1 percent of the sample of full-time working men were haredi.
64
education system to the Israeli public. This finding points to the need to subject the public education system that purports to provide universal schooling at a uniform level of quality to thorough research. 8.
Conclusion
Inequality in the labor market, measured in terms of the labor wage per hour worked of full-time employed men, widened by 32 percent between the early 1980s and the 1990s. In view of the tendency of inequality indices to change at a snail’s pace, this is a significant increase. The rise in inequality is the result of a change in the structure of the return on human capital and an increase in inequality among individuals who have the same reported traits. Hardly any of the upturn traces to greater inequality in ownership of various indicators of human capital. Wages of persons with higher schooling (thirteen years of study or more) climbed steeply relative to those of the poorly schooled. In the early 1980s, workers with sixteen or more years of schooling earned 62 percent more than those with very scanty schooling; in some years in the 1990s, the gap widened to 83 percent. Managers are another group that has benefited from a dramatic increase in relative wage. In early 1980s, they earned 26 percent more than unskilled workers who had the same measured measurable traits (including schooling); this rate escalated to a peak of 57 percent in 1995. One-fourth to one-half of the wage gap is explained by variables that represent different shades of human capital. The empty half of the glass is that
65
the causes of much of the wage disparity are still shrouded in fog. This study shows that inequality among “equals” is also important in explaining the longterm trajectory of labor-market inequality; this factor explains much of the increase in inequality in the 1990s relative to the early 1980s. In the United States, the upturn in inequality among “equals” was believed to trace to an increase in demand for occupational credentials; some have applied this explanation to the Israeli case as well. In regard to Israel, however, this result may reflect a widening of gaps in schooling quality. The increase in return on human capital among the highly schooled (13+ years) coincided with an increase in the proportion of persons with higher schooling. This is an indication of a more rapid escalation of demand for human-capital-intensive workers. To some extent, one may construe increase in the unexplained fraction as further evidence of swift growth (relative to supply) of demand for occupational credentials. This study does not identify the factors that caused demand for schooled (and perhaps skilled) workers to outpace the expansion that occurred in supply of the highly skilled. However, several factors mentioned in this survey receive indirect support. Labor-market inequality increased concurrently with the upturn in inflation in the mid-1980s and contracted immediately after the Economic Stabilization Program stanched inflation successfully. In the late 1980s, inequality did not return to its level at the beginning of the decade
66
because the unemployment rate rose steeply. In the 1990s, inequality climbed to a record level as the labor market absorbed mass immigration.
67
Bibliography Artstein, Y., and S. Sussman (1991), “A Dynamic of Wage Increases in Israel: Market Forces and Inter-Industry Comparisons,” Bank of Israel, Research Department, Discussion Article 91.02. Awwad, Yasser (1998), “Influence of Macroeconomic Variables on Poverty and Income Distribution in Israel,” National Insurance Institute, Annual Survey 1997/98. Central Bureau of Statistics (1997), “Influence of In-Kind Transfers on Income Distribution,” Publication 1049, March. Dahan, Momi (1995), “Influence of Macroeconomic Variables on Income Distribution in Israel,” Bank of Israel Survey, 69:21–48. — and E. Ben-Porat (1997), “The Immigration Wave and IncomeDistribution Inequality,” Economics Quarterly, Year 44, November. Flug, K. and Kasir, N. (1993), “Compliance with the Minimum Wage Law in the Business Sector,” Bank of Israel, Research Department, Discussion Paper 94.12. Gabai, Yosef (1996), “Influence of Fiscal Policy on Income-Distribution Inequality in Israel, Economics Quarterly, Year 43, April. Leviathan, O. (1982), “Development of the Cost-of-Living Allowance Arrangement and Other Wage Clauses,” Economics Quarterly, 115 (December): 753–943. National Insurance Institute (1998), Annual Survey 1997/98.
68
Sussman, Z, and D. Zakkai (1983), “Changes in Wage Structure in General Government during the Rising Inflation Period 1974–1981,” Bank of Israel, Research Department. — and — (1997), “Paradox: Large Supply of Immigrant Doctors and Steep Increase in Nonimmigrants’ Wages, 1990 to 1995,” Bank of Israel, Research Department, Discussion Article 97.06. Yaniv, G. (1986), “Noncompliance with the Minimum Wage [Law]: Theoretical Aspects and Implications for the Israeli Economy,” National Insurance Institute, Research and Planning Division, Study No. 35.
English Sources Bjorklund, A. (1991). “Unemployment and Income Distribution: TimeSeries Evidence from Sweden,” Scandinavian Journal of Economics, 93(3): 457–465. Blejer, M. I. and I. Guerrero (1990). “The Impact of Macroeconomic Policies on Income Distribution: An Empirical Study of the Philippines,” The Review of Economics and Statistics, 72 (August): 414–423. Blinder, A. S. and H. Y. Esaki (1978). “Macroeconomic Activity and Income Distribution in the Postwar United States,” The Review of Economics and Statistics, 60 (November): 604–609.
69
Cook, J. Philip and Robert H. Frank (1995). The Winner-Take-All Society: Why the Few at the Top Get So Much. New York, London and Toronto: Simon and Schuster, Free Press, Martin Kessler Books. Deininger, K. And Squire L. (1996). “Measuring Income Inequality: A New Data Base,” The World Bank Economic Review, 10: 565–91. Eckstein, Zvi and Ron Shachar (1995). “On the Transition to Work of New Immigrants: 1990–92.” Jerusalem: The Maurice Falk Institute for Economic Research in Israel, Discussion Paper No. 95.04. Eckstein, Zvi and Yoram Weiss (1998a). “The Absorption of Highly Skilled Immigrants: Israel, 1990–1995,” mimeo. Eckstein, Zvi and Yoram Weiss (1998b) “The Integration of Immigrants from the Former Soviet Union in the Israeli Labor Market,” mimeo. Flug, Karnit, Nitsa Kasir and Gur Ofer (1992). “The Absorption of Soviet Immigrants into the Labor Market from 1990 Onwards: Aspects of Occupational Substitution and Retention.” Jerusalem: Bank of Israel Discussion Paper No. 92.13. Friedberg, Rachel (1992). “You Can’t Take It With You? Immigrant Assimilation and Portability of Human Capital: Evidence from Israel.” Jerusalem: The Maurice Falk Institute for Economic Research in Israel, Discussion Paper No. 95.02.
70
Gottschalk Peter and Timothy M. Smeeding (1997). “Cross-National Comparison of Earning and Income Inequality,” Journal of Economic Literature, 35 (June): 633–687. IADB (Inter-American Development Bank), Facing Up To Inequality in Latin America, Economic and Social Progress in Latin America 1998– 1999 Report. Juhn, Chinhui, Kevin Murphy and Brooks Pierce (1993). “Wage Inequality and the Rise in the Return to Skills,” Journal of Political Economy, 101: 410–442. Klinov, Ruth, (1999). “Changes in the Wage Structure—An Overview: Israel 1972–1995,” mimeograph. Nolan, B. (1988-89). “Macroeconomic Conditions and the Size Distribution of Income: Evidence from the United Kingdom,” Journal of Post Keynesian Economics, 11 (Winter): 196–221. Silber, J. and B. Z. Zilberfarb (1994). “The Effect of Anticipated and Unanticipated Inflation on Income Distribution: The Israeli Case,” Journal of Income Distribution, 4 (1): 41–50. Summers R. and A. Heston (1994). “Penn World Table (Mark 5.6).” [[?]] Weiss, Yoram and Menachem Gotlibovski (1995). “Immigration, Search and Loss of Skill.” Tel Aviv: Foerder Institute Working Paper 34–95.
71
Text for Figures Figure 15.1. Inequality in Israel, 1965-1997 (*) Gini index Gini index for gross-income inequality among employee-headed households
Figure 15.2. Inequality in the Labor Market, 1980–1996
Figure 15.3. Unexplained Portion of Labor-Market Inequality: The Contribution of Immigration Excl. immigrant dummy variables Incl. immigrant dummy variables
Figure 15.4. Inequality in the Labor Market: Nonimmigrants Only
Figure 15.5. Effect of Ownership and Return Structure of Human Capital: Nonimmigrants Only Quantities Prices